Stories and social structure A structural perspective on literature in society Sociologists as well as scholars of literature study structures: social structures and text structures respectively. Of course, the social and cultural sides of life are felt to be related in some way or another. Influenced by cultural anthropology, sociologists have incorporated culture in their theories. Students of literature have paid attention to social reality in authors’ biographies and the like. The sociology of literature in particular has focused on the relation between society and literature, usually employing the metaphor that literature mirrors society (see Desan et al. 1989 for a synthesis). In this paper, I will pursue the relation between literature and society from a different angle. Instead of society as a whole, I will focus on social behaviour at the individual level first. The stories that people tell about themselves, their social relations, and behaviour are closely connected. From literary and non-literary narratives people derive structural models for the stories of their lives as well as scripts for social action. This explains why narrative structures resemble group structures, opening new perspectives on the analysis of narrative structures as well as the significance of stories to social behaviour. However, the link between stories and larger social formations is at stake too, since the process of defining identities and structuring relationships operates not just at the individual level but also at the level of social groups and organisations. Balance in stories and social groups Stories and social structure seem to be different worlds, studied in different scientific disciplines. Why would anyone assume similarities between narrative and social structure? My argument starts with the observation that a social-psychological theory, which is explicitly relational, applies to affective relations within small groups as well as stories. I will 1 use this theory to introduce a relational perspective, which is relevant to the analysis of literary criticism. In the 1940s Fritz Heider formulated the basic tenet of what was later called balance theory (Heider 1946). It may be summarised as “a person feels uncomfortable if he disagrees on a topic with someone he likes”. This situation is called unbalanced and induces a person to change his affections or opinions. Balance theory predicts that human beings will strive for balanced situations. Soon, this theory was extended to affect relations between three people and more, stating that “a person feels uncomfortable if he dislikes his friends’ friends or if he likes his friends’ enemies”. Group structures are hypothesised to display a tendency towards balance. In a large number of psychological and sociological research projects balance theory has been tested and found to be fairly accurate (cf. Taylor 1970 for an overview). Small groups, e.g. children in an educational setting, often display tendencies towards balance. Also the theory was applied to social relations that are not affect relations in settings that do not study the classic social-psychological small group, e.g. in political science and history. Important to my argument is the application of balance theory to tales (Auster 1980) and drama (Harary 1963 and 1966). Auster and Harary coded the relations between dramatis personae in tale, opera, or play as either positive or negative: like, help versus dislike, combat et cetera. Analysing these relations, they found that the stories and plays always end in complete balance and that most stories and plays contain markedly unbalanced episodes. They concluded that these narratives display a development towards balance as predicted by balance theory. Social-psychological balance theory seems to be a theory of the happy ending story too. In 1956 Cartwright and Harary (Cartwright and Harary 1956) used mathematical graph theory, which deals with structures that we commonly refer to as networks, to formalise balance theory. They represented actors by nodes or vertices and affect relations between actors by arcs that were either solid, representing positive affect, or dashed, representing 2 negative affect. Figure 1 represents a balanced network of affect relations. Person B likes person A but he dislikes C, who dislikes A. So B dislikes his friend’s enemy as predicted by balance theory. Cartwright and Harary proofed that balance in a network of affect relations may easily be detected. For example, each closed sequence or circle of affect relations in the network that contains zero or an even number of negative affects, is balanced. A C B D Figure 1 - A balanced network of affections. The formalisation had two major effects. On the one hand, balance theory could now easily be applied to any structure that may be represented as a network of positive and negative arcs. For example countries at war or peace, competing or co-operating firms, agreement and disagreement between politicians, et cetera. This has lead to application of balance theory on a wide range of topics. On the other hand, formalisation stimulated the search for alternative models. Network analysts contributed substantially to the theory’s 3 development by extending it to hierarchical group structures. Probably balance theory is the first social theory that owes its development in large part to network analysis (Manhart 1995). There is no need to explain all of these models here (see Wasserman, S., & Faust, K. (1994) for an excellent summary of this theory and the associated mathematical models). However, it is important to note that balance theory as well as methods of network analysis are fundamentally relational. Since I will introduce a relational paradigm in a later section, it is sensible to elaborate a little on the relational character of network analysis as opposed to statistics first. Network analysis versus statistics New theories develop in relation with new methods or methodologies. Statistics has fuelled social research and theory-formation for a long time. Of late network analysis has emerged as an alternative method. Although network analysis may use statistical concepts, it is based on radically different assumptions. Let us contrast the two methods. Statistics focuses on entities’ attributes and their statistical relations. Consider the network of affections depicted in Figure 1. In a statistical approach, a variable ‘social group’ would be defined with values that may be labelled ‘hipsters’ and ‘squares’ in case of kids at high school. Each youngster would then be assigned to one of either categories, e.g. pupils A and B would be considered ‘hip’. Other attributes may be measured and compared to the first. For example ‘hip’ may correlate with a specific social background. If ‘hipsters’ tend to have other social backgrounds than ‘squares’, the attributes ‘social group’ and ‘social background’ are said to be associated statistically. Note that units of analysis are always aggregated in statistics: the set of hip pupils is characterised on average by a specific social background. So it is not necessarily true that each hipster has the specific social background. In contrast, network analysis does not measure an entity’s attributes, but (social) relations between entities. Network analysis looks for patterns of relations. In Fig. 1, for example, A and B would be clustered because they are related by a positive tie, whereas they 4 have negative ties to and from C and D. Within network analysis, a group of friends is delineated by their pattern of affect relations, instead of their individual characteristics or the kind of person they claim to be. This characterises network analysis as a relational or structural method. There is no need to specify beforehand the number of clusters nor their labels (‘hip’ versus ‘square’) as required in statistics. Clusters are extracted from social relations at the individual level. They are interpreted and labelled afterwards. Also data do not need to be aggregated. This is very important to the analysis of narratives. Since the number of characters in a story uses to be fairly small, statistical aggregation makes no sense. Network analysis, however, may be applied since it analyses the relations between characters. The capacity to analyse socially constructed classifications is an important feature of network analysis that is central to my argumentation. As indicated before, I will concentrate on balance theory although this is but one of the theories developed in social network analysis. Social network analysis is a method that is being used in a wide area of research, ranging from chemistry, logistics and diffusion processes (Rogers 1979) to cultural anthropology (Hage and Harary 1983) and content analysis (Carley 1993). It formalises a relational way of thinking – thinking in relations. Let us now turn to a relational paradigm in sociological theory. A relational paradigm In 1992 Harrison C. White published a book titled Identity and Control: A Structural Theory of Social Action. Sociologists hailed it as an important but difficult book (Boudon 1993; Calhoun 1993; Meyer 1993; Tilly 1993). Interestingly, White stresses the importance of stories as elements that structure social reality. “Stories come from and become a medium for control efforts: that is the core” White writes (White 1992: 68), and he advises to measure social relations by listening to stories. A social theory dealing with stories may well help to explain the relation between social structure and narrative structure that we are interested in. So let us have a closer look at it. 5 To understand White’s theory, I argue, it is necessary to realise that White grounds it on a relational paradigm. According to White, an identity – an ‘I’ or a ‘we’ – arises from an effort at control. If there are two entities that establish a relation, because they try to control one another, a sense of identity comes into being. White replaces individualistic self- awareness ‘cogito ergo sum’ by social awareness: ‘there is an effort at control, so there is an I (or we), and a you’.